The present invention relates in general to the field of vehicle/robot localization, and more particularly, to a method and an apparatus for uncertainty modeling of a point cloud.
For a driving vehicle to detect the environment using laser scanner(s) and to use the acquired data in tasks like localization, navigation and recognition, etc., it is important to model the uncertainty of the point cloud for both the map data and real time observed data. The uncertainty model of the point cloud allows us to calculate the probability of matching or recognition, and a good uncertainty model may help to make the calculation more accurate and more efficient. Prior methods include: ICP algorithm;[1] IDC algorithm;[2] NURB (spline surface in 3D);[3] PAMI ICP algorithm;[4] 2D-NDT;[5] 3D-NDT;[6] Region Growing NDT.[7]